bge-reranker-v2-m3
Open Weights
BAAI
BAAI's Apache-licensed multilingual reranker, a 0.6B cross-encoder built on the bge-m3 base. Unlike an embedding model it takes a query and a passage together and outputs a relevance score directly, which can be mapped to [0,1] with a sigmoid. It is the lightweight recommendation in BAAI's own reranker lineup and pairs naturally with the bge-m3 embeddings already in this catalog.
Strengths
- Apache-2.0 licensed and fully self-hostable
- Strong multilingual coverage, BAAI's recommended choice for multilingual reranking
- Only 0.6B parameters, with fp16 inference for extra speed
- Outputs a relevance score directly rather than an embedding to compare
- Pairs with the bge-m3 embedding model it is built on
Caveats
- 512-token maximum input - long passages must be chunked before reranking
- A reranker, not a generative model; it needs a first-stage retriever
- BAAI publishes its evaluation results as charts rather than figures, so no scores are recorded here
- No release date is stated on the model card
- Vendor
- BAAI
- Context Window
- 512 tokens
- License
- Apache-2.0
- Modalities
- text
Capabilities
Vision
Audio
Video
Tool Use
Resources
Pricing
Open weights under Apache-2.0 - self-host at no licence cost. At 0.6B it runs on modest hardware, and fp16 inference is supported.